OpenAI’s limited preview of its GPT-5.6 model family is the second instance in recent weeks in which a frontier AI model’s rollout has been shaped by U.S. government review before wider release. This pattern raises questions about how quickly Australian developers and businesses can expect access to leading AI systems going forward.
OpenAI began previewing GPT-5.6, released in three tiers named Sol, Terra and Luna, to approximately 20 vetted partner organizations this week, sharing the list of participants with the U.S. government ahead of launch. The restriction followed a June 2 executive order directing federal agencies to develop a benchmarking and evaluation framework for new frontier AI models, according to OpenAI’s own announcement. OpenAI has not set a general availability date, saying only that broader access is expected “in the coming weeks.”
The gating mechanism differs from the export control directive that led Anthropic to suspend global access to its Claude Fable 5 and Mythos 5 models on June 12, a restriction that applied to all users outside the U.S. ( including Australia) and was lifted July 1 after roughly 19 days.
OpenAI’s GPT-5.6 review is a government-coordinated preview process rather than a formal export control action, but the practical effect for users outside the initial partner group, including in Australia, is similar.
Neither OpenAI nor the U.S. government has indicated whether any Australian organizations are among the roughly 20 partner organizations granted preview access to GPT-5.6. OpenAI chief executive Sam Altman said the company complied with the government’s request but does not believe gated access should become standard practice for future model releases.
Australia has no formal mechanism for negotiating early or guaranteed access to frontier AI models released by U.S. companies, unlike its role in agreements such as AUKUS covering defence technology cooperation. Access to commercial AI models has generally followed each company’s own staged rollout timeline, a process that has functioned similarly to a global default rather than a negotiated arrangement specific to Australia or other allied nations.
Industry observers have not yet assessed whether the pattern of U.S. government-gated AI releases, now evident in two separate cases within weeks of each other, will become a recurring feature of frontier AI rollouts or remains tied to the specific security reviews triggered in each case. Fable 5’s restriction was resolved after Anthropic implemented updated safety classifiers addressing a specific reported vulnerability. The basis on which GPT-5.6’s broader release will be judged ready has not been publicly detailed by OpenAI or the U.S. government.
